Kelly, W. | Search bioRxiv

bioRxiv2022

Examining the efficacy of localised gemcitabine therapy for the treatment of pancreatic cancer using a hybrid agent-based model

The prognosis for pancreatic ductal adenocarcinoma (PDAC) patients has not significantly improved in the past 3 decades, highlighting the need for more effective treatment approaches. Poor patient outcomes and lack of response to therapy can be attributed, in part, to the dense, fibrotic nature of PDAC tumours, which impedes the uptake of systemically administered drugs. Wet-spun alginate fibres loaded with the chemotherapeutic agent gemcitabine have been developed as a potential tool for overcoming the physical and biological barriers presented by the PDAC tumour microenvironment and deliver high concentrations of drug to the tumour directly over an extended period of time. While exciting, the practicality, safety, and effectiveness of these devices in a clinical setting requires further investigation. Furthermore, an in-depth assessment of the drug-release rate from these devices needs to be undertaken to determine whether an optimal release profile exists. Using a hybrid computational model (agent-based model and partial differential equation system), we developed a simulation of pancreatic tumour growth and response to treatment with gemcitabine loaded alginate fibres. The model was calibrated using in vitro and in vivo data and simulated using a finite volume method discretization. We then used the model to compare different intratumoural implantation protocols and gemcitabine-release rates. In our model, the primary driver of pancreatic tumour growth was the rate of tumour cell division and degree of extracellular matrix deposition. We were able to demonstrate that intratumoural placement of gemcitabine loaded fibres was more effective than peritumoural placement. Additionally, we found that an exponential gemcitabine release rate would improve the tumour response to fibres placed peritumourally. Altogether, the model developed here is a tool that can be used to investigate other drug delivery devices to improve the arsenal of treatments available for PDAC and other difficult-to-treat cancers in the future. Author SummaryPancreatic cancer has a dismal prognosis with a median survival of 3-5 months for untreated disease. The treatment of pancreatic cancer is challenging due to the dense nature of pancreatic tumours which impedes retention of drug at the tumour site. As such, systemic administration of chemotherapies, such as gemcitabine, has a limited efficacy. To overcome this, sustained-release devices have been proposed. These devices are injected locally and release drug slowly over time, providing a concentrated local, sustained drug concentration. To investigate the possible efficacy of these devices, we developed a mathematical model that would allow us to probe treatment perturbations in silico. We modelled the individual cancer cells and their growth and death from gemcitabine loaded into the sustained delivery devices. Our platform allows future investigations for these devices to be run in silico so that we may better understand the forms of the drug release-profile that are necessary for optimal treatment.

cancer biology↗

bioRxiv2019

Learning supervised embeddings for large scale sequence comparisons

Similarity-based search of sequence collections is a core task in bioinformatics, one dominated for most of the genomic era by exact and heuristic alignment-based algorithms. However, even efficient heuristics such as BLAST may not scale to the data sets now emerging, motivating a range of alignment-free alternatives exploiting the underlying lexical structure of each sequence.\n\nIn this paper, we introduce SuperVec, a novel supervised approach to learning sequence embeddings. Our method extends earlier Representation Learning (RL) based methods to include jointly contextual and class-related information for each sequence during training. This ensures that related sequence fragments have proximal representations in the target space, better reflecting the structure of the domain.\n\nSuch representations may be used for downstream machine learning tasks or employed directly. Here, we apply SuperVec embeddings to a sequence retrieval task, where the goal is to retrieve sequences with the same family label as a given query. The SuperVec approach is extended further through H-SuperVec, a tree-based hierarchical method which learns embeddings across a range of feature spaces based on the class labels and their exclusive and exhaustive subsets.\n\nExperiments show that supervised learning of embeddings based on sequence labels using SuperVec and H-SuperVec provides a substantial improvement in retrieval performance over existing (unsupervised) RL-based approaches. Further, the new methods are an order of magnitude faster than BLAST for the database retrieval task, supporting hybrid approaches in which SuperVec rapidly filters the collection so that only potentially relevant records remain, allowing slower, more accurate methods to be executed quickly over a far smaller dataset. Thus, we may achieve faster query processing and higher precision than before.\n\nFinally, for some problems, direct use of embeddings is already sufficient to yield high levels of precision and recall. Extending this work to encompass weaker homology is the subject of ongoing research.

bioinformatics↗